MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Working on fairly unrelated stuff, I needed to prove the following, fairly easy results, and I wonder if anyone can provide references to the literature. Not being a probabilist I wouldn't know where to start looking.

I. Let $X$ and $Y$ be two integrable random variables, and let $X'$ and $Y'$ have the same distributions, respectively, and be indepdendent from $X$ and $Y$. Then $$ E[\min \{X,X'\} + \min \{Y,Y'\} - 2\min\{ X,Y'\}] \geq 0 $$ and equality holds iff $X$ and $Y$ have the same distribution

II. Let $X$ be a random variable, and let $\cal F$ be some sub-algebra. Let $Y$ have the same conditional distribution as $X$ over $\cal F$, indepndently from $X$ over $\cal F$, and let $Z$ have the same distribution as $X$, independently from $X$. Then $$ E[\min \{X,Y\}] \geq E[ \min \{X,Z\} ] $$ and equality holds iff $X$ is independent from $\cal F$.

share|cite|improve this question
I would suggest using functional notation for $\min,$ that is $\min(X, Y),$ otherwise the equations are very confusing. – Igor Rivin Sep 10 '12 at 21:53

If the cumulative distribution function of $X$ is $G$ while the probabiity density $g=G^\prime,$ then the probability density of $\min(X, X^\prime)$ is $2 g G.$ Similarly, if the CDF of $Y$ is $H,$ with density $h,$ the probability density of $\min(Y, Y^\prime)$ is $2 h H,$ and that of $\min(X, Y')$ is $g H + h G.$ So, your expectation is $2\int x (g G + h H - g H - h G) d x = \int x d(G-H)^2.$ If $G=H,$ the integral is $0,$ as you conjecture, but if $G\neq H$ the expectation is nonnegative if the random variable is nonnegative.

I don't quite understand the second question...

share|cite|improve this answer
Thank you, but I do know how to prove these things - my question is whether this exists in the literature. – Itaï BEN YAACOV Sep 11 '12 at 7:04
To explain II, maybe a quick proof is best: You can reduce fairly easily to the case where $X$ is the indicator function of some event $A$. Let $f = P[A|{\cal F}]$. Then the inequality becomes $E[f^2] \geq E[f]^2$, i.e., $V(f) \geq 0$, and equality holds iff $V(f) = 0$, i.e., if $f$ is constantly $E[f] = P[A]$, i.e., if $A$ is independent from $\cal F$. – Itaï BEN YAACOV Sep 11 '12 at 7:58
@Itai: re your first comment: I have a certain sense of deja-vu, so it probably does exist in the literature; looking it up is harder than proving it :( – Igor Rivin Sep 11 '12 at 12:23

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.